A flame image feature dimension reduction quantitative identification method, device and equipment

Abstract

The present application relates to the technical field of model identification and artificial intelligence, and particularly relates to a flame image feature dimension reduction quantitative identification method, device and equipment. The method comprises the following steps: extracting multi-dimensional features of historical flame images; analyzing the correlation between the features, performing first dimension reduction on the multi-dimensional features according to the correlation, and obtaining a first feature set; calculating principal components of each feature in the first feature set and feature values of each principal component, and taking the principal components of each feature as a second feature set; calculating the contribution rate of each principal component in the second feature set according to the feature values, performing second dimension reduction on the second feature set according to the contribution rate, and obtaining a third feature set; calculating a comprehensive score of the historical flame image according to each principal component in the third feature set and the contribution rate of each principal component; and constructing a quantitative identification model according to the comprehensive score and a fire source parameter of the historical flame image. Through the embodiment, a quantitative identification model is constructed by simplifying the image feature vector dimension, and quantitative identification of the flame image is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of model identification and artificial intelligence, and in particular to a flame image feature dimension reduction quantitative identification method, device and equipment. BACKGROUND

[0002] At present, the traditional fire monitoring mainly relies on temperature, smoke and photosensitive fire detectors to judge the development of fire accidents through the threshold of the detector. However, for some large space environment (warehouse, factory, forest, etc.) fire scene, the temperature change, gas product, heat radiation and other produced by fire burning need to be detected by sensors after a long distance and a complex way, so the traditional fire detector has certain disadvantages. At the same time, the traditional detection method is affected by the accuracy of the detector and the environmental factors (light, haze, etc.), so that the accuracy is not high, the false alarm rate is high, and the unnecessary waste of human resources is caused.

[0003] The prior art (patent publication No. CN108052950A) uses principal component analysis method to reduce the dimension of two-dimensional matrix of flame image, and then maps the selected principal component back to RGB image space through image reconstruction technology for verification, but the principal component analysis on the original data of flame image has the problems of large amount of calculation and loss of original effective information of picture, and the features obtained after dimension reduction are directly applied to the working condition classification of flame, the essence is still the traditional qualitative classification, which leads to low precision of the method for classifying flame features, and certain limitations.

[0004] Now there is an urgent need for a flame image feature dimension reduction quantitative identification method to solve the problems of large amount of calculation in the prior art method for analyzing flame image, low classification precision of traditional qualitative classification method, and certain limitations. SUMMARY

[0005] To solve the problems in the prior art, the embodiments of the present application provide a flame image feature dimension reduction quantitative identification method, device and equipment, which considers the image feature and the fire source state parameter set at the time of fire, and constructs a quantitative function model by simplifying the image feature vector dimension, thereby realizing the quantitative identification of flame image.

[0006] In order to solve the above technical problems, the specific technical scheme of the present application is as follows:

[0007] On the one hand, the embodiments of the present application provide a flame image feature dimension reduction quantitative identification method, which comprises,

[0008] extracting multi-dimensional features of historical flame images;

[0009] analyze the correlation between each feature in the multi-dimensional features, perform first dimension reduction on the multi-dimensional features according to the correlation, and obtain a first feature set;

[0010] calculate principal components of each feature in the first feature set and feature values of each principal component, and take the principal components of each feature as a second feature set;

[0011] calculate the contribution rate of each principal component in the second feature set according to the feature values, perform second dimension reduction on the second feature set according to the contribution rate, and obtain a third feature set;

[0012] calculate a comprehensive score of the historical flame image according to each principal component in the third feature set and the contribution rate of each principal component;

[0013] construct a quantitative recognition model according to the comprehensive score and the fire source parameters of the historical flame image;

[0014] extract target features of a to-be-recognized flame image according to the categories of each feature in the first feature set, so as to calculate the fire source parameters of the to-be-recognized flame image according to the target features and the quantitative recognition model, and predict the trend of the to-be-recognized flame according to the fire source parameters of the to-be-recognized flame image.

[0015] Further, the first dimension reduction on the multi-dimensional features according to the correlation to obtain the first feature set further comprises,

[0016] determine a plurality of features whose correlation exceeds a first predetermined threshold value, and obtain a plurality of feature groups;

[0017] store any one feature in the feature groups into the first feature set.

[0018] Further, the formula for calculating the contribution rate of each principal component in the second feature set according to the feature values is,

[0019]

[0020] wherein η i represents the contribution rate of the i th principal component, λ i represents the feature value of the i th principal component, and m represents the number of principal components in the second feature set.

[0021] Further, the second dimension reduction on the second feature set according to the contribution rate to obtain the third feature set further comprises,

[0022] sort each principal component in the second feature set according to the contribution rate from large to small;

[0023] According to the ordering, the contribution rates of the principal components are sequentially added, and after each new contribution rate is added, it is determined whether the sum of the added contribution rates exceeds a second predetermined threshold value; if not, the contribution rate of the next principal component is added according to the ordering; if yes, the principal component corresponding to the added contribution rates is stored in the third feature set.

[0024] Further, calculating the comprehensive score of the historical flame image according to the principal components in the third feature set and the contribution rates of the principal components further comprises,

[0025] According to the principal components in the third feature set and the features in the first feature set, a dimension reduction result of the historical flame image is calculated.

[0026] The product of each dimension of the dimension reduction result and the corresponding contribution rate is calculated to obtain a plurality of products.

[0027] The sum of each product is calculated to obtain the comprehensive score of the historical flame image.

[0028] Further, the step of calculating the fire source parameter of the to-be-identified flame image according to the target feature and the quantitative identification model comprises,

[0029] The principal components of the target feature and the feature values corresponding to the principal components are calculated, and the principal components of the target feature are taken as the first principal component feature set of the to-be-identified flame image.

[0030] The contribution rates of the principal components in the first principal component feature set are calculated according to the feature values, and the second dimension reduction is performed on the first principal component feature set according to the contribution rates to obtain a second principal component feature set.

[0031] The comprehensive score of the to-be-identified flame image is calculated according to the principal components in the second principal component feature set and the contribution rates of the principal components.

[0032] The fire source parameter of the to-be-identified flame image is calculated according to the comprehensive score of the to-be-identified flame image and the quantitative identification model.

[0033] Further, the multi-dimensional features include flame area, flame perimeter, flame contrast, entropy, flame correlation, flame energy and / or flame fractal dimension.

[0034] The formula for calculating the flame area is,

[0035]

[0036] wherein A fwherein M represents the flame width, N represents the flame length, b(i, j) represents the pixel of the flame image after the grayscale and binary processing, and i and j represent the coordinates of the pixel;

[0037] The formula for calculating the flame perimeter is

[0038]

[0039] wherein Ln represents the flame perimeter, Num() is a statistical quantity function, i = 1, 3, 5, 7 represents searching the next pixel point in the clockwise direction from an arbitrary point on the flame contour, if the next pixel point is found in the oblique direction, the statistical quantity is 1, i = 2, 4, 6, 8 represents the next pixel point is found in the horizontal or vertical direction, and n represents the total number of points of the boundary of the flame contour;

[0040] The formula for calculating the flame contrast is

[0041]

[0042] wherein σ con represents the flame contrast, and i and j represent the coordinates of the pixel;

[0043] The formula for calculating the entropy is

[0044]

[0045] wherein σ ent represents the entropy;

[0046] The formula for calculating the flame correlation degree is

[0047]

[0048] wherein σ cor represents the correlation degree;

[0049]

[0050]

[0051] The formula for calculating the flame energy is

[0052]

[0053] wherein σ ener represents the flame energy;

[0054] The formula for calculating the flame fractal dimension is

[0055]

[0056] wherein FD represents the flame fractal dimension, r represents the length of a side of a small cube, and Nr represents the number of small cubes obtained by covering the flame image with the small cubes, the small cube representing a part of the flame image.

[0057] In another aspect, the embodiments also provide a device for dimension reduction and quantitative identification of flame image features, comprising,

[0058] a multi-dimension feature extraction unit configured to extract multi-dimension features of historical flame images;

[0059] a first dimension reduction unit configured to analyze the correlation between each two features in the multi-dimension features, perform first dimension reduction on the multi-dimension features according to the correlation, and obtain a first feature set;

[0060] a principal component calculation unit configured to calculate the principal components of each feature in the first feature set and the eigenvalues of each principal component, and take the principal components of each feature as a second feature set;

[0061] a second dimension reduction unit configured to calculate the contribution rates of each principal component in the second feature set according to the eigenvalues, perform second dimension reduction on the second feature set according to the contribution rates, and obtain a third feature set;

[0062] a comprehensive score calculation unit configured to calculate the comprehensive scores of the historical flame images according to each principal component in the third feature set and the contribution rates of each principal component;

[0063] a quantitative identification model construction unit configured to construct a quantitative identification model according to the comprehensive scores and the fire source parameters of the historical flame images;

[0064] a to-be-identified fire source parameter calculation unit configured to extract target features of a to-be-identified flame image according to the categories of each feature in the first feature set, calculate the fire source parameters of the to-be-identified flame image according to the target features and the quantitative identification model, and predict the trend of the to-be-identified flame according to the fire source parameters of the to-be-identified flame image.

[0065] In another aspect, the embodiments also provide a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above method.

[0066] In another aspect, the embodiments also provide a computer readable storage medium, which stores a computer program, wherein the computer program is executable on a processor to implement the above method.

[0067] With the embodiments herein, firstly, multi-dimension features of historical flame images are extracted. Since the number of multi-dimension features is large, if the principal components of each multi-dimension feature are directly calculated, the calculation amount will be huge. Therefore, before the principal components of multi-dimension features are calculated, the correlation between each two features in the multi-dimension features is analyzed, the multi-dimension features are first reduced in dimension according to the correlation, to obtain a first feature set, so that the correlated features are removed, the feature dimension is preliminarily reduced, and the calculation amount of the principal components is reduced. Then, the principal components of each feature in the first feature set and the characteristic values of the principal components are calculated, the principal components of each feature are taken as a second feature set, that is, the dimension of the second feature set is the same as that of the first feature set, but the data is different. Then, the contribution rates of the principal components in the second feature set are calculated according to the characteristic values, the second feature set is second reduced in dimension according to the contribution rates, to obtain a third feature set, that is, the third feature set includes part of the principal components of the second feature set. Then, the comprehensive scores of the historical flame images are calculated according to the principal components in the third feature set and the contribution rates of the principal components, so that the calculation amount is further reduced. Then, the comprehensive scores and the fire source parameters of the historical flame images are fitted, a quantitative function relationship between the comprehensive scores and the fire source parameters is established, and a quantitative recognition model is obtained, so as to calculate the fire source parameters of a flame image to be recognized by using the quantitative recognition model. The quantitative recognition of the flame image is realized, and the problems of large calculation amount in the prior art, low classification precision of the traditional qualitative classification method, and certain limitation are solved. BRIEF DESCRIPTION OF DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the embodiments herein or in the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments herein, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0069] Figure 1 It is a schematic diagram of an implementation system of the dimension reduction quantitative recognition method for flame image features in the embodiments herein.

[0070] Figure 2 It is a flowchart of the dimension reduction quantitative recognition method for flame image features in the embodiments herein.

[0071] Figure 3 It is a process of first reducing the multi-dimension features in dimension according to the correlation to obtain a first feature set in the embodiments herein.

[0072] Figure 4 It is a process of calculating the comprehensive scores of the historical flame images according to the principal components in the third feature set and the contribution rates of the principal components in the embodiments herein.

[0073] Figure 5 The step of calculating the fire source parameter of the to-be-identified flame image according to the target feature and the quantitative identification model is shown in the embodiments of the present application;

[0074] Figure 6 The structural schematic diagram of the dimension reduction quantitative identification device for the flame image feature in the embodiments of the present application is shown;

[0075] Figure 7 The structural schematic diagram of the computer device in the embodiments of the present application is shown.

[0076]

Explanation of reference numerals

[0077] 101, terminal;

[0078] 102, server;

[0079] 601, multi-dimensional feature extraction unit;

[0080] 602, first dimension reduction unit;

[0081] 603, principal component calculation unit;

[0082] 604, second dimension reduction unit;

[0083] 605, comprehensive score calculation unit;

[0084] 606, quantitative identification model construction unit;

[0085] 607, to-be-identified fire source parameter calculation unit;

[0086] 702, computer device;

[0087] 704, processing device;

[0088] 706, storage resource;

[0089] 708, driving mechanism;

[0090] 710, input / output module;

[0091] 712, input device;

[0092] 714, output device;

[0093] 716, presentation device;

[0094] 718, graphical user interface;

[0095] 720, network interface;

[0096] 722, communication link;

[0097] 724, communication bus. Detailed Implementation

[0098] The technical solutions in the embodiments described below will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments described herein, and not all of the embodiments. Based on the embodiments described herein, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this document.

[0099] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings herein are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0100] like Figure 1 The diagram illustrates a system implementation of a dimensionality reduction and quantitative identification method for flame image features according to an embodiment of the present invention. The system may include a terminal 101 and a server 102. A communication connection is established between the terminal 101 and the server 102, enabling data interaction. Operators can input historical flame images and corresponding fire source parameters to the server 102 via the terminal 101. The server 102 constructs a quantitative identification model based on the historical flame images. Then, the server 102 receives the flame image to be identified input from the terminal 101, calculates the fire source parameters of the flame image using the constructed quantitative identification model, and finally sends the calculated fire source parameters back to the terminal 101. This allows operators to predict the trend of the flame to be identified based on the fire source parameters of the flame image.

[0101] In the embodiments of this specification, the server 102 may be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0102] In an optional embodiment, the terminal 101 can include, but is not limited to, electronic devices such as desktop computers, tablets, notebooks, etc. Optionally, the operating system running on the electronic device can include, but is not limited to, Android, IOS, Linux, Windows, etc.

[0103] In addition, it should be noted that, Figure 1 The application environment shown is only one application environment provided by the present disclosure, and other application environments can also be included in actual applications, and the present specification is not limited.

[0104] In order to solve the problems in the prior art, the embodiments of the present application provide a flame image feature dimension reduction quantitative identification method, device and equipment, which considers the image feature and the fire source state parameter set at the time of fire, and constructs a quantitative function model by simplifying the image feature vector dimension, thereby realizing quantitative identification of the flame image. Figure 2 The flowchart of the flame image feature dimension reduction quantitative identification method is shown. In this figure, the process of quantitatively identifying the fire source parameters is described, but more or fewer operation steps can be included based on conventional or non-inventive labor. The order of the steps listed in the embodiments is only one of the many execution orders, and does not represent the only execution order. In actual system or device product execution, the method order shown in the embodiments or the drawings can be executed in sequence or in parallel. Specifically, as Figure 2 As shown, the method can be executed by a processing unit on a server, and the method can include:

[0105] Step 201: Extracting multi-dimensional features of historical flame images;

[0106] Step 202: Analyzing the correlation between each feature in the multi-dimensional features, performing first dimension reduction on the multi-dimensional features according to the correlation, and obtaining a first feature set;

[0107] Step 203: Calculating the principal components of each feature in the first feature set and the characteristic values of each principal component, and taking the principal components of each feature as a second feature set;

[0108] Step 204: Calculating the contribution rate of each principal component in the second feature set according to the characteristic values, performing second dimension reduction on the second feature set according to the contribution rate, and obtaining a third feature set;

[0109] Step 205: Calculating the comprehensive score of the historical flame image according to each principal component in the third feature set and the contribution rate of each principal component;

[0110] Step 206: Constructing a quantitative identification model according to the comprehensive score and the fire source parameters of the historical flame image;

[0111] Step 207: extracting target features of the to-be-identified flame image according to the categories of each feature in the first feature set, so as to calculate the fire source parameter of the to-be-identified flame image according to the target features and the quantitative identification model, and predict the trend of the to-be-identified flame according to the fire source parameter of the to-be-identified flame image.

[0112] By the method described in the embodiments, first, the multi-dimensional features of the historical flame image are extracted. Since the number of features of the multi-dimensional features is large, if the principal components of each multi-dimensional feature are directly calculated, the calculation amount will be huge. Therefore, before calculating the principal components of the multi-dimensional features, the correlation between each two features in the multi-dimensional features is analyzed, the multi-dimensional features are first reduced in dimension according to the correlation, and a first feature set is obtained, so as to remove the related features and preliminarily reduce the feature dimension, thereby reducing the calculation amount of the principal components. Then, the principal components of each feature in the first feature set and the characteristic values of each principal component are calculated, the principal components of each feature are taken as a second feature set, that is, the dimension of the second feature set is the same as that of the first feature set, but the data is different. Then, the contribution rates of each principal component in the second feature set are calculated according to the characteristic values, the second feature set is second reduced in dimension according to the contribution rates, and a third feature set is obtained, that is, the third feature set includes part of the principal components of the second feature set. Then, the comprehensive score of the historical flame image is calculated according to each principal component in the third feature set and the contribution rate of each principal component, so as to further reduce the calculation amount. Then, the comprehensive score and the fire source parameter of the historical flame image are fitted, a quantitative function relationship between the comprehensive score and the fire source parameter is established, and a quantitative identification model is obtained, so as to calculate the fire source parameter of the to-be-identified flame image by using the quantitative identification model. The quantitative identification of the flame image is realized, and the problems of large calculation amount in the method for analyzing the flame image in the prior art, low classification accuracy of the traditional qualitative classification method, and certain limitations are solved.

[0113] In the embodiments, the flame image of the fire scene can be acquired by an image acquisition device, and the original image data is obtained. Then, the original flame image is processed by using the weighted average method to obtain a flame gray image, and the flame gray image is processed by using the binary method to obtain a flame binary image. Finally, the multi-dimensional features of the flame binary image and / or the flame binary image are extracted.

[0114] Specifically, according to one embodiment, the flame multi-dimensional features can include but are not limited to flame area, flame perimeter, flame contrast, entropy, flame correlation, flame energy and / or flame fractal dimension.

[0115] The flame area is one of the most intuitive indicators representing the scale of the flame. The flame area can be obtained by counting the number of pixels representing the flame in the binary image. The formula for calculating the flame area is (1):

[0116]

[0117] wherein A f represents the flame area, M represents the flame width, N represents the flame length, b(i,j) represents the pixel of the flame image after the gray scale and binary processing, i and j represent the coordinates of the pixel;

[0118] The flame perimeter refers to the contour length of separating the flame from the background. The contour perimeter is calculated by using chain code, that is, the chain code of the flame contour is calculated first, and then the odd and even numbers of the chain code are counted to obtain the perimeter of the flame contour. The specific steps are as follows:

[0119] Step 1: selecting an arbitrary point on the flame contour as a starting point, and recording the coordinates;

[0120] Step 2: searching the next adjacent pixel point from the direction coded as 1 in a clockwise direction, and recording the code;

[0121] Step 3: repeating the above steps until returning to the starting point, that is, obtaining a set of chain codes.

[0122] It can be seen that for a known flame image, the selection of the starting point will not affect the calculation result of the perimeter.

[0123] The formula for calculating the flame perimeter is (2):

[0124]

[0125] wherein Ln represents the flame perimeter, Num() is a statistical quantity function, i=1,3,5,7 represents searching the next pixel point in a clockwise direction from an arbitrary point on the flame contour, if the next pixel point is found in the inclined direction, the statistical quantity is 1, i=2,4,6,8 represents that the next pixel point is found in the horizontal or vertical direction, and n represents the total number of points of the boundary of the flame contour.

[0126] The gray-level co-occurrence matrix (GLCM) is used to calculate the texture features of the flame image. The gray-level co-occurrence matrix refers to obtaining a co-occurrence matrix P(i,j) by the following formula (3) and formula (4), and then extracting some characteristic parameters of the matrix to quantify the texture of the image, so as to reflect the comprehensive information of the image gray scale about direction, adjacent interval, change amplitude, etc.

[0127] f(x+a,y+b)=j|{(x,y),(x+a,y+b)∈M×N} (3)

[0128] In the gray level co-occurrence matrix, any point (x, y) in the image (N x N) and another point (x + a, y + b) deviated from it, a and b represent the specified deviation position.

[0129]

[0130] x and y refer to the horizontal row number and vertical column number of the picture pixel, i.e. the coordinate position of the pixel.

[0131] The flame contrast reflects the degree of local change in the image, representing the clarity and depth of the texture of the image. The greater the contrast, the deeper the texture groove, the clearer the visual, and the formula for calculating the flame contrast is (5):

[0132]

[0133] wherein σ con represents the flame contrast, and i and j represent the coordinates of the pixel;

[0134] Entropy represents the non-uniformity of the texture in the flame image. When all elements in the co-occurrence matrix have the maximum randomness, the greater the dispersion, the greater the entropy. The formula for calculating the entropy is (6)

[0135]

[0136] wherein σ ent represents the entropy;

[0137] The flame correlation degree refers to the similarity of the elements of the gray level co-occurrence matrix in the row or column direction, reflecting the correlation of the local gray scale in the image. When the matrix element values are uniform and equal, the correlation degree is the largest. The formula for calculating the flame correlation degree is (7)

[0138]

[0139] wherein σ cor represents the correlation degree;

[0140]

[0141]

[0142]

[0143]

[0144] The flame energy is the sum of the squares of all elements of the gray level co-occurrence matrix, reflecting the uniformity of the image gray scale distribution and the fineness of the texture. The greater the energy value, the more uniform the texture, and the more regular the change. The formula for calculating the flame energy is (12):

[0145]

[0146] Where, σ ener This indicates the energy of the flame;

[0147] The fractal dimension of a flame reflects the effectiveness of the space occupied by the complex shape of the flame profile and is a measure of the irregularity of the flame profile. The formula for calculating the fractal dimension of a flame is (13).

[0148]

[0149] Where FD represents the fractal dimension of the flame, r represents the length of one side of the small cube, and N r This indicates the number of small cubes obtained by covering the flame image with the small cubes, where each small cube represents a portion of the flame image.

[0150] It should be noted that the small cube is an existing concept in the fractal dimension of flame, without any actual physical meaning. It simply means that the flame image is uniformly divided into small cubes with a fixed side length of r, and the total number of small cubes covering the flame is N. r By changing the value of r, you can obtain many r and N values. r The correspondence is given. The value of the flame fractal dimension FD can be calculated using the above formula (13).

[0151] According to one embodiment of this article, such as Figure 3 As shown, the process of performing a first dimensionality reduction on the multi-dimensional features based on the correlation to obtain a first feature set further includes,

[0152] Step 301: Identify multiple features whose correlation exceeds a first predetermined threshold, and obtain multiple feature groups;

[0153] Step 302: Store any one feature from the feature group into the first feature set.

[0154] In the embodiments described herein, a higher correlation indicates a stronger positive correlation and more similar trends between features. The upper limit of the correlation is 1. If the correlation exceeds a first predetermined threshold (e.g., 90%, 95%, or 98%), any feature in the feature group is stored in the first feature set. It should be noted that the correlation between two or more dimensions can be calculated, and this embodiment does not impose any limitations.

[0155] This can be understood as, through Figure 3 The method shown initially reduces the dimensionality of multi-dimensional image features, thereby reducing the computational load of subsequent steps.

[0156] Then, principal component analysis is used to calculate the principal components and eigenvalues ​​of each feature in the first feature set, and the principal components of each feature are used as the second feature set.

[0157] In this embodiment, the size of each principal component is 1×m, where m represents the number of features in the first feature set. The principal components can be expressed as formula (14):

[0158] Pk = [a 1,k ,a 2,k ,a 3,k ,...,a s,k (14)

[0159] Where Pk represents the k-th principal component, a s,k Let represent the component of the k-th principal component Pk, and s represent the number of features in the first feature set.

[0160] Each principal component corresponds to an eigenvalue. The first feature set includes 5 features (flame area A). f Flame circumference Ln, flame contrast σ con Entropy σ ent correlation with flame σ cor For example, λ1 represents the eigenvalue of principal component P1, λ2 represents the eigenvalue of principal component P2, λ3 represents the eigenvalue of principal component P3, λ4 represents the eigenvalue of principal component P4, and λ5 represents the eigenvalue of principal component P5.

[0161] Then, the contribution rate of each principal component is calculated using formula (15):

[0162]

[0163] Where, η i λ represents the contribution rate of the i-th principal component. i Let m represent the eigenvalue of the i-th principal component, and m represent the number of principal components in the second feature set, which is the same as the number of features in the first feature set.

[0164] The contribution rates η1, η2, η3, η4, and η5 of principal component P1, P2, P3, P4, and P5 can be obtained using formula (15).

[0165] Then, a second dimensionality reduction is performed on the second feature set based on the contribution rate to obtain the third feature set. For example, the principal components with the largest contribution rates in the second feature set are retained, thereby reducing the computational load in subsequent steps.

[0166] According to one embodiment of the present disclosure, in order to make the third feature set obtained after the second dimension reduction still maintain the properties of the original feature set, thereby ensuring the calculation accuracy, the step of performing the second dimension reduction on the second feature set according to the contribution rates to obtain the third feature set further includes,

[0167] sorting the principal components in the second feature set according to the order from large to small of the contribution rates;

[0168] adding the contribution rates of the principal components in turn according to the sorting, and determining whether the sum of the added contribution rates exceeds a second predetermined threshold value after each time a contribution rate is added; if not, continuing to add the contribution rate of the next principal component according to the sorting; if yes, storing the principal component corresponding to the added contribution rates into the third feature set.

[0169] In the embodiments of the present disclosure, the second predetermined threshold value can be set according to artificial experience, for example, 95%, and the contribution rates of the principal components are added in turn according to the sorting order from large to small, and after each time a contribution rate is added, it is determined whether the sum of the added contribution rates exceeds the second predetermined threshold value, if not, the contribution rate of the next principal component is added according to the sorting, if yes, the principal component corresponding to the added contribution rates is stored into the third feature set. It can be understood that by selecting the principal components in the third feature set, the dimension of the original image feature can be reduced to the number of principal components in the third feature set, thereby reducing the calculation amount of the subsequent steps.

[0170] For example, the contribution rates of the principal components P1, P2, P3, P4 and P5 in the second feature set are η1, η3, η2, η4 and η5 in the order from large to small, and each contribution rate is added in turn according to the order of η1+η3+η2+η4+η5, and after each time a contribution rate is added, it is determined whether the sum of the added contribution rates exceeds the second predetermined threshold value. If the sum of η1, η3 and η2 is greater than the second predetermined threshold value, the principal components P1, P2 and P3 are stored into the third feature set.

[0171] Then, the comprehensive score of the historical flame image is calculated according to the principal components in the third feature set and the contribution rates of the principal components. Specifically, according to one embodiment of the present disclosure, as shown in Figure 4 the process of calculating the comprehensive score of the historical flame image according to the principal components in the third feature set and the contribution rates of the principal components further includes,

[0172] Step 401: calculating the dimension reduction result of the historical flame image according to the principal components in the third feature set and the features in the first feature set;

[0173] Step 402: calculating the product of each dimension of the dimension reduction result and the corresponding contribution rate to obtain a plurality of products;

[0174] Step 403: summing each of the products to obtain a comprehensive score of the historical flame image.

[0175] In the embodiments herein, the formula for calculating the dimension-reduced result of the historical flame image according to the principal components in the third feature set and each feature in the first feature set is (16):

[0176]

[0177] wherein Y represents the dimension-reduced result of the historical flame image, including n dimensions, wherein n is the same as the number of principal components in the third feature set, Pn represents the nth principal component, t1-tn represent the contribution rates of the first to nth principal components, and s represents the number of features in the first feature set. s s features in the first feature set.

[0178] Taking the third feature set including principal components P1, P2 and P3, and the first feature set including 5 features (flame area A f , flame perimeter Ln, flame contrast σ con , entropy σ ent and flame correlation σ cor ) as an example, the dimension-reduced result Y of the historical flame image is (17):

[0179]

[0180] Then, the product of each dimension of the dimension-reduced result Y and the corresponding contribution rate is calculated to obtain n products, and finally the sum of each product is calculated to obtain a comprehensive score of the historical flame image. In combination with the dimension-reduced result Y shown in formula (16), formula (18) can be obtained:

[0181]

[0182] wherein F represents the comprehensive score.

[0183] Taking the third feature set including principal components P1, P2 and P3, and the first feature set including 5 features (flame area A f , flame perimeter Ln, flame contrast σ con , entropy σ ent and flame correlation σ cor ) and the dimension-reduced result Y of the historical flame image shown in formula (17) as an example, the comprehensive score Y can be obtained as (19):

[0184]

[0185] Finally, a quantitative identification model is constructed according to the comprehensive score F and the fire source parameters of the historical flame image. The fire source parameters can be obtained by measuring the historical flame image with a measuring instrument.

[0186] Exemplarily, the fire source parameters corresponding to all historical flame images, such as the burning speed v and the flame temperature T, are respectively constructed into a quantitative function relationship with the comprehensive score F. The construction method is as follows: a data fitting method is used, which can be fitted into a linear function relationship, a power exponential function relationship, a logarithmic function relationship, an exponential function relationship, and the like, so as to establish a quantitative function relationship between the low-dimensional vector comprehensive score and the fire source parameters. For example, the relationship between the comprehensive score F and the burning speed v is obtained through fitting as follows:

[0187] v = function(F);

[0188] The relationship between the comprehensive score F and the flame temperature T is obtained through fitting as follows:

[0189] T = function(F).

[0190] Therefore, the fire source parameters can be estimated by using images through the method of the embodiment, and quantitative identification is achieved.

[0191] According to one embodiment of the present application, as shown in Figure 5 the step of calculating the fire source parameters of the to-be-identified flame image according to the target features and the quantitative identification model comprises,

[0192] Step 501: calculating principal components of the target features and feature values corresponding to each principal component, and taking the principal components of the target features as a first principal component feature set of the to-be-identified flame image;

[0193] Step 502: calculating a contribution rate of each principal component in the first principal component feature set according to the feature values, and performing the second dimension reduction on the first principal component feature set according to the contribution rate to obtain a second principal component feature set;

[0194] Step 503: calculating a comprehensive score of the to-be-identified flame image according to each principal component in the second principal component feature set and the contribution rate of each principal component;

[0195] Step 504: calculating the fire source parameters of the to-be-identified flame image according to the comprehensive score of the to-be-identified flame image and the quantitative identification model.

[0196] In the embodiment of the present application, the target features are extracted from the to-be-identified flame image according to the categories of each feature in the first feature set. Then, the fire source parameters of the to-be-identified flame image can be calculated according to the method as shown in Figure 5 , so as to predict the trend of the to-be-identified flame according to the calculated fire source parameters.

[0197] Based on the same inventive concept, the embodiment of the present application further provides a device for dimension reduction and quantitative identification of flame image features, as shown in the figure, comprising, Figure 6

[0198] a multi-dimension feature extraction unit 601, configured to extract multi-dimension features of historical flame images;

[0199] a first dimension reduction unit 602, configured to analyze the correlation between each two features in the multi-dimension features, perform first dimension reduction on the multi-dimension features according to the correlation, and obtain a first feature set;

[0200] a principal component calculation unit 603, configured to calculate the principal components of each feature in the first feature set and the feature values of each principal component, and take the principal components of each feature as a second feature set;

[0201] a second dimension reduction unit 604, configured to calculate the contribution rates of each principal component in the second feature set according to the feature values, perform second dimension reduction on the second feature set according to the contribution rates, and obtain a third feature set;

[0202] a comprehensive score calculation unit 605, configured to calculate the comprehensive scores of the historical flame images according to each principal component in the third feature set and the contribution rates of each principal component;

[0203] a quantitative identification model construction unit 606, configured to construct a quantitative identification model according to the comprehensive scores and the fire source parameters of the historical flame images;

[0204] a to-be-identified fire source parameter calculation unit 607, configured to extract target features of a to-be-identified flame image according to the categories of each feature in the first feature set, so as to calculate the fire source parameters of the to-be-identified flame image according to the target features and the quantitative identification model, and predict the trend of the to-be-identified flame according to the fire source parameters of the to-be-identified flame image.

[0205] Since the principle of the above device for solving the problem is similar to the above method, the implementation of the above device can refer to the implementation of the above method, and the repeated parts will not be described here.

[0206] As shown in the figure, Figure 7 ​A structural diagram of a computer device according to an embodiment of the present application is shown. The device in the present application can be a computer device in the present embodiment, which executes the method of the present application described above. The computer device 702 can include one or more processing devices 704, such as one or more central processing units (CPUs), each of which can implement one or more hardware threads. The computer device 702 can also include any storage resources 706 for storing any kind of information, such as code, settings, data, etc. Without limitation, for example, the storage resources 706 can include any one or combination of the following: any type of RAM, any type of ROM, a flash memory device, a hard disk, an optical disk, etc. More generally, any storage resource can store information using any technology. Further, any storage resource can provide volatile or non-volatile retention of information. Further, any storage resource can represent a fixed or removable component of the computer device 702. In one case, the computer device 702 can perform any operation of the associated instructions when the processing device 704 executes the associated instructions stored in any storage resource or combination of storage resources. The computer device 702 also includes one or more drive mechanisms 708, such as a hard disk drive mechanism, an optical disk drive mechanism, etc., for interacting with any storage resources.

[0207] The computer device 702 can also include an input / output module 710 (I / O) for receiving various inputs (via input devices 712) and for providing various outputs (via output devices 714). One particular output mechanism can include a presentation device 716 and an associated graphical user interface (GUI) 718. In other embodiments, the input / output module 710 (I / O), the input devices 712, and the output devices 714 can also not be included, just as a computer device in a network. The computer device 702 can also include one or more network interfaces 720 for exchanging data with other devices via one or more communication links 722. One or more communication buses 724 couple the above-described components together.

[0208] The communication links 722 can be implemented in any manner, for example, through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication links 722 can include any combination of hardwired links, wireless links, routers, gateway functionality, name servers, etc., governed by any protocol or combination of protocols.

[0209] Corresponding to the method of Figures 2-5 The present embodiment also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to perform the above steps.

[0210] The embodiments herein also provide a computer readable instruction, wherein the program in the computer readable instruction, when executed by a processor, causes the processor to perform the method as shown in Figures 2-5

[0211] It should be understood that the size of the serial number of the processes described above in the various embodiments herein does not mean the order of execution, and the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments herein.

[0212] It should also be understood that in the embodiments herein, the term "and / or" is only a description of the association relationship of the associated objects, which means that there can be three relationships. For example, A and / or B can represent three cases of A alone, A and B together, and B alone. In addition, the character " / " in this paper generally represents that the front and rear associated objects are in an "or" relationship.

[0213] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in the above description in general terms. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this paper.

[0214] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0215] In several embodiments provided herein, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displays or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can also be electrical, mechanical or other forms of connection.

[0216] ​The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e. may be located in one place, or may be distributed on multiple network units. Part or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0217] In addition, each functional unit in each embodiment herein can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0218] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions herein, essentially or the part of the prior art that contributes to the technical solutions, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment herein. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0219] The principles and implementation manners of the present application are described in the specific embodiments herein, and the above embodiment descriptions are only used to help understand the methods and core ideas thereof; meanwhile, for those skilled in the art, according to the ideas herein, the specific implementation manners and application ranges will have changes, and the above description of the present application should not be understood as a limitation.

Claims

1. A method for dimension reduction quantitative identification of flame image features, characterized in that, The method comprises the steps of: extracting multi-dimensional features of historical flame images; analyzing correlations between features in the multi-dimensional features, performing first dimension reduction on the multi-dimensional features according to the correlations to obtain a first feature set; calculating principal components of each feature in the first feature set and feature values of each principal component, and taking the principal components of each feature as a second feature set; calculating contribution rates of each principal component in the second feature set according to the feature values, performing second dimension reduction on the second feature set according to the contribution rates to obtain a third feature set; calculating a comprehensive score of the historical flame image according to each principal component in the third feature set and the contribution rate of each principal component; constructing a quantitative recognition model according to the comprehensive score and a fire source parameter of the historical flame image; extracting target features of a to-be-recognized flame image according to categories of each feature in the first feature set, so as to calculate a fire source parameter of the to-be-recognized flame image according to the target features and the quantitative recognition model, and predict a trend of the to-be-recognized flame according to the fire source parameter of the to-be-recognized flame image; constructing a quantitative recognition model according to the comprehensive score and a fire source parameter of the historical flame image comprises: using a data fitting method to establish a quantitative function relationship between the comprehensive score and the fire source parameter.

2. The method of claim 1, wherein, The method further comprises the steps of: determining a plurality of features with a correlation exceeding a first predetermined threshold value to obtain a plurality of feature groups; storing any one feature in the feature groups into the first feature set.

3. The method of claim 1, wherein, The formula for calculating the contribution rate of each principal component in the second feature set according to the feature values is: wherein η i represents the contribution rate of the i-th principal component, λ i represents the eigenvalue of the i-th principal component, and m represents the number of principal components in the second feature set.

4. The method of claim 1, wherein, The method further comprises the steps of: sorting each principal component in the second feature set in descending order of the contribution rate; adding the contribution rates of the principal components in sequence according to the sorting, and determining whether the sum of the added contribution rates exceeds a second predetermined threshold value after each new contribution rate is added; if not, continue adding the contribution rate of the next principal component according to the sorting; if yes, store the principal component corresponding to the added contribution rates into the third feature set.

5. The method of claim 1, wherein, The method further comprises the steps of: calculating a dimension reduction result of the historical flame image according to the principal components in the third feature set and each feature in the first feature set; calculating a product of each dimension of the dimension reduction result and the corresponding contribution rate to obtain a plurality of products; summing each product to obtain the comprehensive score of the historical flame image.

6. The method of claim 1, wherein, The method further comprises the steps of: calculating principal components of the target features and feature values corresponding to each principal component, and taking the principal components of the target features as a first principal component feature set of the to-be-recognized flame image; According to the characteristic value, a contribution rate of each principal component in the first principal component feature set is calculated, and the first principal component feature set is subjected to the second dimension reduction according to the contribution rate, so as to obtain a second principal component feature set; According to each principal component in the second principal component feature set and the contribution rate of each principal component, a comprehensive score of the to-be-identified flame image is calculated; According to the comprehensive score of the to-be-identified flame image and the quantitative identification model, a fire source parameter of the to-be-identified flame image is calculated.

7. The method of claim 1, wherein, The multi-dimension features include a flame area, a flame perimeter, a flame contrast, an entropy, a flame correlation degree, a flame energy and / or a flame fractal dimension; A formula for calculating the flame area is, where A f represents the flame area, M represents the flame width, N represents the flame length, b(i,j) represents the pixel of the flame image after the grayscale and binary processing, and i and j represent the coordinates of the pixel. A formula for calculating the flame perimeter is, wherein, Ln represents the flame perimeter, Num() is a statistical quantity function, i=1, 3, 5, 7 represents that a next pixel point is searched in a clockwise direction from an arbitrary point on a flame contour, if the next pixel point is found in a tilt direction, the statistical quantity is 1, i=2, 4, 6, 8 represents that the next pixel point is found in a horizontal or vertical direction, and n represents a total point number of a boundary of the flame contour; A formula for calculating the flame contrast is, where σ con represents the flame contrast, i and j represent the coordinates of the pixel; A formula for calculating the entropy is, wherein σ ent denotes the entropy; A formula for calculating the flame correlation degree is, wherein σ cor denotes the correlation degree; A formula for calculating the flame energy is, where σ ener represents the flame energy; A formula for calculating the flame fractal dimension is, wherein FD represents the flame fractal dimension, r represents the length of a side of a small cube, N r represents the number of small cubes obtained by covering the flame image with the small cubes, the small cubes representing part of the flame image.

8. A device for dimensionally reducing and quantitatively identifying flame image features, characterized by comprises, A multi-dimension feature extraction unit is configured to extract multi-dimension features of historical flame images; A first dimension reduction unit is configured to analyze a correlation between each two features in the multi-dimension features, and perform a first dimension reduction on the multi-dimension features according to the correlation, so as to obtain a first feature set; A principal component calculation unit is configured to calculate principal components of each feature in the first feature set and characteristic values of each principal component, and take the principal components of each feature as a second feature set; A second dimension reduction unit is configured to calculate a contribution rate of each principal component in the second feature set according to the characteristic values, and perform a second dimension reduction on the second feature set according to the contribution rate, so as to obtain a third feature set; A comprehensive score calculation unit is configured to calculate a comprehensive score of the historical flame image according to each principal component in the third feature set and the contribution rate of each principal component; A quantitative identification model construction unit is configured to construct a quantitative identification model according to the comprehensive score and a fire source parameter of the historical flame image; A to-be-identified fire source parameter calculation unit is configured to extract target features of a to-be-identified flame image according to a category of each feature in the first feature set, so as to calculate a fire source parameter of the to-be-identified flame image according to the target features and the quantitative identification model, and predict a trend of the to-be-identified flame according to the fire source parameter of the to-be-identified flame image. The method of constructing the quantitative identification model according to the comprehensive score and the fire source parameter of the historical flame image comprises: A data fitting method is adopted to establish a quantitative function relationship between the comprehensive score and the fire source parameter.

9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the method of any one of claims 1 to 7.

Images